| Abstract |
We first analyze advance payment guarantees to estimate firm probabilities of default and loss given default. We use multiple discriminant, logit, and survival analyses to estimate the default risk. The probability of default is determined by guarantee characteristics, such as the guarantee size and the period. Firm characteristics, such as firm size, credit rating, and financial ratios, also influence the default risk. Default forecasting based on these models has much greater forecasting power than models based on the credit rating. Second, we estimate loss given default based on the censored least squares method, nonlinear regressions, and a decision tree model. We find that considering macroeconomic factors during the guarantee period is essential for estimations of loss given default. Among the three noted methodologies, the decision tree model best predicts loss given default. |